Potential for Improving Among‐Reader Precision and Accuracy of Walleye Age Estimates with Minimal Training. (6th June 2019)

• Record Type: Journal Article
• Authors: Dembkowski, Daniel J.; Isermann, Daniel A.; Koenigs, Ryan P.
• Is Part Of: North American journal of fisheries management. Volume 39:Number 4(2019)
• Page Start: 625
• Page End: 636
• Publication Date: 2019-06-06
• Subjects: Fishery management -- United States -- Periodicals; 333.956097305
• DOI: 10.1002/nafm.10293
• Languages: English
• ISSNs: 0275-5947

Abstract

Abstract: Within most resource management agencies, fish ages assigned by multiple readers are used to estimate age‐based population metrics and to develop state or regional growth standards under the assumption that among‐reader precision and accuracy are high. A cursory evaluation suggested that precision of age estimates among seven individuals who routinely estimate Walleye Sander vitreus age for the Wisconsin Department of Natural Resources was remarkably low (otolith mean coefficient of variation [CV] = 37%; dorsal spine mean CV = 35%), which prompted concern and interest in whether among‐reader precision could be improved with a minimal level of training. Consequently, we compared among‐reader precision and accuracy before and after a 1‐d training exercise. We distributed images of sectioned otoliths and sectioned dorsal spines from a random sample of 50 Walleye, along with images of structures from 25 known‐age Walleye, to 21 readers grouped into beginner, intermediate, and advanced experience levels based on responses to a pretraining survey. Among‐reader precision was substantially higher after training (otolith mean CV = 16%; dorsal spine mean CV = 15%) than before (otolith mean CV = 27%; dorsal spine mean CV = 26%). Accuracy of age estimates also improved after training, but greater improvements were observed for otoliths (mean difference between estimated and known ages before training = 0.80 year; after training = 0.15 year) than for dorsal spines (mean difference between estimated and known ages before training = 0.38 year; after training = 0.22 year). Similar improvements in precision and accuracy were evident for all experience levels. Our results suggest that a simple, low‐cost age estimation training can substantially increase precision and accuracy of age estimates among a large group of readers. However, additional training and quality control measures may be required to achieve higher levels of precision and accuracy.